Table 1 and Additional File 2: Table S1 show the number of T2D cases and controls in the European, African, and Hispanic/Latino populations, for which we had sufficient sample sizes to evaluate the performance of the PRS, across the 8 eMERGE sites. When the target population was European, the trans-ancestry PRS constructed by PRS-CSx explained 9.2% of the variation in the T2D status on the liability scale, with an AUC of 0.66 and OR/SD of 1.96 (95% confidence interval [CI] 1.91–2.02), after adjusting for age, sex, top 10 genetic PCs and eMERGE study sites (Table 2; Additional File 2: Table S3). The PRS provided predictive power above and beyond the covariates, increasing the AUC from 0.74 (the covariates-only model) to 0.79 when combining covariates and the PRS as predictors (Additional File 2: Table S3). To assess the clinical utility of the trans-ancestry PRS, we compared the risk of T2D among individuals in progressively more extreme cutoffs of the PRS distribution relative to the rest of the samples. Individuals of European ancestry in the top decile of the trans-ancestry PRS had
